You are viewing a plain text version of this content. The canonical link for it is here.
Posted to solr-user@lucene.apache.org by Bernd Fehling <be...@uni-bielefeld.de> on 2017/07/25 11:47:51 UTC
SolrCloud wildcard query result order change
With SolrCloud 6.4.2 (5 shards on 5 server) and a wildcard query
I get different results between the same query.
I assume this is alltogether due to the distributed search and
the response time of each server and the constant score of 1.0 ???
Is there any config where I can set the shard order (shard_1...shard_n)
as result order for results with same score?
Or for results of wildcard query?
Any ideas?
Regards
Bernd
Re: SolrCloud wildcard query result order change
Posted by Susheel Kumar <su...@gmail.com>.
What is the query you are executing if you can share. Due you think
difference could be due to updates/ingestion happening same time?
Thanks,
Susheel
On Tue, Jul 25, 2017 at 7:47 AM, Bernd Fehling <
bernd.fehling@uni-bielefeld.de> wrote:
> With SolrCloud 6.4.2 (5 shards on 5 server) and a wildcard query
> I get different results between the same query.
>
> I assume this is alltogether due to the distributed search and
> the response time of each server and the constant score of 1.0 ???
>
> Is there any config where I can set the shard order (shard_1...shard_n)
> as result order for results with same score?
>
> Or for results of wildcard query?
>
> Any ideas?
>
> Regards
> Bernd
>
Re: SolrCloud wildcard query result order change
Posted by Erick Erickson <er...@gmail.com>.
When scores are identical, the tie is broken by the _internal_ Lucene doc ID.
For the same doc on two different replicas of the same shard, the internal
ID is not only different, but two docs may be ordered (by internal doc ID) one
way on replica1 and reversed on replica2.
To guarantee identical ordering I'd just sort by ID, or maybe date (or
whatever).
If you sort by date, you still potentially have the same problem, two docs may
have an identical date so to insure ordering always have your last sort by
your <uniqueKey>, e.g. sort=date asc,id asc
Best,
Erick
On Tue, Jul 25, 2017 at 6:22 AM, Bernd Fehling
<be...@uni-bielefeld.de> wrote:
> Am 25.07.2017 um 15:09 schrieb Mikhail Khludnev:
>> Since all scores are equal it just can not break this tie. Add id as a sort
>> clause to make results deterministic.
>
> What about setting statsCache to ExactSharedStatsCache?
> But there is all the same score 1.0, so maybe not a solution?
>
> Bernd
>
>>
>> On Tue, Jul 25, 2017 at 3:39 PM, Bernd Fehling <
>> bernd.fehling@uni-bielefeld.de> wrote:
>>
>>> Any wildcard query will do it, e.g. .../select?q=ant*&wt=json&...
>>>
>>> A couple of "shift + reload" (to bypass cache) in the browser and you
>>> will see that the order of the result changes sometimes.
>>>
>>> Definately no updates/ingestion because it's currently a
>>> SolrCloud test system with only 12 mio. docs.
>>>
>>> Regards
>>> Bernd
>>>
>>>
>>> Am 25.07.2017 um 14:20 schrieb Susheel Kumar:
>>>> What is the query you are executing if you can share. Due you think
>>>> difference could be due to updates/ingestion happening same time?
>>>>
>>>> Thanks,
>>>> Susheel
>>>>
>>>> On Tue, Jul 25, 2017 at 7:47 AM, Bernd Fehling <
>>>> bernd.fehling@uni-bielefeld.de> wrote:
>>>>
>>>>> With SolrCloud 6.4.2 (5 shards on 5 server) and a wildcard query
>>>>> I get different results between the same query.
>>>>>
>>>>> I assume this is alltogether due to the distributed search and
>>>>> the response time of each server and the constant score of 1.0 ???
>>>>>
>>>>> Is there any config where I can set the shard order (shard_1...shard_n)
>>>>> as result order for results with same score?
>>>>>
>>>>> Or for results of wildcard query?
>>>>>
>>>>> Any ideas?
>>>>>
>>>>> Regards
>>>>> Bernd
>>>>>
>>>>
Re: SolrCloud wildcard query result order change
Posted by Bernd Fehling <be...@uni-bielefeld.de>.
Am 25.07.2017 um 15:09 schrieb Mikhail Khludnev:
> Since all scores are equal it just can not break this tie. Add id as a sort
> clause to make results deterministic.
What about setting statsCache to ExactSharedStatsCache?
But there is all the same score 1.0, so maybe not a solution?
Bernd
>
> On Tue, Jul 25, 2017 at 3:39 PM, Bernd Fehling <
> bernd.fehling@uni-bielefeld.de> wrote:
>
>> Any wildcard query will do it, e.g. .../select?q=ant*&wt=json&...
>>
>> A couple of "shift + reload" (to bypass cache) in the browser and you
>> will see that the order of the result changes sometimes.
>>
>> Definately no updates/ingestion because it's currently a
>> SolrCloud test system with only 12 mio. docs.
>>
>> Regards
>> Bernd
>>
>>
>> Am 25.07.2017 um 14:20 schrieb Susheel Kumar:
>>> What is the query you are executing if you can share. Due you think
>>> difference could be due to updates/ingestion happening same time?
>>>
>>> Thanks,
>>> Susheel
>>>
>>> On Tue, Jul 25, 2017 at 7:47 AM, Bernd Fehling <
>>> bernd.fehling@uni-bielefeld.de> wrote:
>>>
>>>> With SolrCloud 6.4.2 (5 shards on 5 server) and a wildcard query
>>>> I get different results between the same query.
>>>>
>>>> I assume this is alltogether due to the distributed search and
>>>> the response time of each server and the constant score of 1.0 ???
>>>>
>>>> Is there any config where I can set the shard order (shard_1...shard_n)
>>>> as result order for results with same score?
>>>>
>>>> Or for results of wildcard query?
>>>>
>>>> Any ideas?
>>>>
>>>> Regards
>>>> Bernd
>>>>
>>>
Re: SolrCloud wildcard query result order change
Posted by Mikhail Khludnev <mk...@apache.org>.
Since all scores are equal it just can not break this tie. Add id as a sort
clause to make results deterministic.
On Tue, Jul 25, 2017 at 3:39 PM, Bernd Fehling <
bernd.fehling@uni-bielefeld.de> wrote:
> Any wildcard query will do it, e.g. .../select?q=ant*&wt=json&...
>
> A couple of "shift + reload" (to bypass cache) in the browser and you
> will see that the order of the result changes sometimes.
>
> Definately no updates/ingestion because it's currently a
> SolrCloud test system with only 12 mio. docs.
>
> Regards
> Bernd
>
>
> Am 25.07.2017 um 14:20 schrieb Susheel Kumar:
> > What is the query you are executing if you can share. Due you think
> > difference could be due to updates/ingestion happening same time?
> >
> > Thanks,
> > Susheel
> >
> > On Tue, Jul 25, 2017 at 7:47 AM, Bernd Fehling <
> > bernd.fehling@uni-bielefeld.de> wrote:
> >
> >> With SolrCloud 6.4.2 (5 shards on 5 server) and a wildcard query
> >> I get different results between the same query.
> >>
> >> I assume this is alltogether due to the distributed search and
> >> the response time of each server and the constant score of 1.0 ???
> >>
> >> Is there any config where I can set the shard order (shard_1...shard_n)
> >> as result order for results with same score?
> >>
> >> Or for results of wildcard query?
> >>
> >> Any ideas?
> >>
> >> Regards
> >> Bernd
> >>
> >
>
>
--
Sincerely yours
Mikhail Khludnev
Re: SolrCloud wildcard query result order change
Posted by Susheel Kumar <su...@gmail.com>.
i thought you said different results i.e. different count.
On Tue, Jul 25, 2017 at 8:39 AM, Bernd Fehling <
bernd.fehling@uni-bielefeld.de> wrote:
> Any wildcard query will do it, e.g. .../select?q=ant*&wt=json&...
>
> A couple of "shift + reload" (to bypass cache) in the browser and you
> will see that the order of the result changes sometimes.
>
> Definately no updates/ingestion because it's currently a
> SolrCloud test system with only 12 mio. docs.
>
> Regards
> Bernd
>
>
> Am 25.07.2017 um 14:20 schrieb Susheel Kumar:
> > What is the query you are executing if you can share. Due you think
> > difference could be due to updates/ingestion happening same time?
> >
> > Thanks,
> > Susheel
> >
> > On Tue, Jul 25, 2017 at 7:47 AM, Bernd Fehling <
> > bernd.fehling@uni-bielefeld.de> wrote:
> >
> >> With SolrCloud 6.4.2 (5 shards on 5 server) and a wildcard query
> >> I get different results between the same query.
> >>
> >> I assume this is alltogether due to the distributed search and
> >> the response time of each server and the constant score of 1.0 ???
> >>
> >> Is there any config where I can set the shard order (shard_1...shard_n)
> >> as result order for results with same score?
> >>
> >> Or for results of wildcard query?
> >>
> >> Any ideas?
> >>
> >> Regards
> >> Bernd
> >>
> >
>
>
Re: SolrCloud wildcard query result order change
Posted by Bernd Fehling <be...@uni-bielefeld.de>.
Any wildcard query will do it, e.g. .../select?q=ant*&wt=json&...
A couple of "shift + reload" (to bypass cache) in the browser and you
will see that the order of the result changes sometimes.
Definately no updates/ingestion because it's currently a
SolrCloud test system with only 12 mio. docs.
Regards
Bernd
Am 25.07.2017 um 14:20 schrieb Susheel Kumar:
> What is the query you are executing if you can share. Due you think
> difference could be due to updates/ingestion happening same time?
>
> Thanks,
> Susheel
>
> On Tue, Jul 25, 2017 at 7:47 AM, Bernd Fehling <
> bernd.fehling@uni-bielefeld.de> wrote:
>
>> With SolrCloud 6.4.2 (5 shards on 5 server) and a wildcard query
>> I get different results between the same query.
>>
>> I assume this is alltogether due to the distributed search and
>> the response time of each server and the constant score of 1.0 ???
>>
>> Is there any config where I can set the shard order (shard_1...shard_n)
>> as result order for results with same score?
>>
>> Or for results of wildcard query?
>>
>> Any ideas?
>>
>> Regards
>> Bernd
>>
>